DeepYardDeepYard
C

CausalFlow

Causal debugging framework that converts failed LLM agent traces into minimal counterfactual fixes

Open SourceFree

About

CausalFlow is an interventional debugging framework designed for LLM agents that performs causal attribution analysis on failed execution traces. When an agent fails, it identifies the exact breakdown points through causal analysis and automatically generates minimal counterfactual repairs—showing what should have happened instead. Ideal for researchers and developers building complex agent systems who need systematic debugging beyond trial-and-error.

Details

Type
Integrations
Language

Tags

observabilityevaluationframeworkautonomousopen-sourcepython